Layer-specific cholinergic control of human and mouse cortical synaptic plasticity

Individual cortical layers have distinct roles in information processing. All layers receive cholinergic inputs from the basal forebrain (BF), which is crucial for cognition. Acetylcholinergic receptors are differentially distributed across cortical layers, and recent evidence suggests that different populations of BF cholinergic neurons may target specific prefrontal cortical (PFC) layers, raising the question of whether cholinergic control of the PFC is layer dependent. Here we address this issue and reveal dendritic mechanisms by which endogenous cholinergic modulation of synaptic plasticity is opposite in superficial and deep layers of both mouse and human neocortex. Our results show that in different cortical layers, spike timing-dependent plasticity is oppositely regulated by the activation of nicotinic acetylcholine receptors (nAChRs) either located on dendrites of principal neurons or on GABAergic interneurons. Thus, layer-specific nAChR expression allows functional layer-specific control of cortical processing and plasticity by the BF cholinergic system, which is evolutionarily conserved from mice to humans.

C ortical acetylcholine (ACh) signalling shapes neuronal circuit development and underlies specific aspects of cognitive functions and behaviors, including attention, learning, memory and motivation [1][2][3][4][5][6] . On the basis of anatomical findings, control of cortical processing by projections from sparse cholinergic nuclei in the basal forebrain (BF) could be much more specific than classically thought 7,8 . Within the mouse BF, a topographic organization exists by which different areas of the medial prefrontal cortex (mPFC) are innervated by different BF cholinergic neurons 7 . Moreover, these neurons preferentially target superficial or deep cortical layers 7 . Both muscarinic and nicotinic ACh receptors (mAChRs and nAChRs) are expressed in a layer-dependent fashion as well [9][10][11] , opening the possibility that cholinergic control of cortical processing is layer specific. Indeed, the distinct, layer-dependent expression of nAChRs in the mPFC could support a layer-dependent control of excitability of pyramidal neurons by cholinergic projections from the BF 11,12 . Applications of ACh show that superficial layer 2/3 (L2/3) pyramidal neurons are inhibited by nAChR activation on interneurons, while deep L6 pyramidal neurons are excited by postsynaptic nAChRs 11,[13][14][15] .
The cellular and sub-cellular location of nAChRs may not only determine how excitability in neuronal circuitries is affected, but may also decide how plasticity of glutamatergic synapses is affected by cholinergic inputs. Activation of nAChRs located on presynaptic terminals can increase glutamate release from synapses 16,17 . In particular, presynaptic nAChRs containing a7 subunits, which have high calcium permeability 18 , can cause long-term potentiation of glutamatergic synapse strength in different brain regions [19][20][21] . In the PFC, despite the expression of a7-containing nAChRs on L5 pyramidal neurons and strong transient modulation of thalamic excitatory inputs by nAChR activation 11,22,23 , nAChRs located on GABAergic interneurons augment inhibitory synaptic transmission and reduce excitability of L5 pyramidal neuron dendrites, thereby suppressing long-term potentiation of glutamatergic synapses 23,24 . In contrast to L5, inhibitory GABAergic and excitatory glutamatergic transmission onto PFC L6 pyramidal neurons are not modulated by nAChRs 11 . Instead, L6 pyramidal neurons express nAChRs themselves and are therefore directly activated by ACh 11,13,14 . These findings suggest that the mechanisms by which nAChRs alter synaptic plasticity of glutamatergic synapses in L5 pyramidal neurons may not be in place in L6.
It is not known whether postsynaptically located nAChRs in the PFC modulate long-term plasticity of glutamatergic synapses and whether this can be induced by endogenous ACh release. Here we find that endogenous ACh release modulates cortical plasticity rules with layer specificity: in L6 endogenous ACh release modulates plasticity in the opposite direction from superficial layers. Endogenous ACh augments long-term strengthening of glutamatergic synapses on L6 pyramidal neurons by activating heteromeric postsynaptic nAChRs containing b2 and a5 subunits. In addition, we find these mechanisms also operate in the human neocortex, where layer-specific expression of functional nAChRs supports opposite cholinergic modulation of synaptic plasticity in superficial and deep cortical layers.

Results
Nicotine facilitates tLTP of L6 pyramidal neuron synapses. Nicotinic ACh receptors located at presynaptic locations can alter synaptic plasticity of glutamatergic synapses 19,23,24 . To test whether postsynaptic nAChRs affect synaptic plasticity of glutamatergic synapses, whole-cell recordings were made from L6 pyramidal neurons of mouse prelimbic (PrL)-mPFC (P21-30) that express nAChRs (Fig. 1a,b). Layer 6 was identified under oblique illumination as a layer containing relatively densely packed neurons with small somata at a perpendicular distance of 4B650 mm from the pia. These neurons generally had regular spiking properties (Fig. 1b) and displayed strong inward currents in response to a local application of ACh (1 mM, 10 ms, in presence of 400 nM atropine, local application protocol I, Methods section) aimed at the cell body (Fig. 1b), as was shown previously [11][12][13][14] .
To investigate the effect of postsynaptic nAChR stimulation on long-term plasticity of glutamatergic synapses, excitatory postsynaptic potentials (EPSPs) were evoked every 7 s by extracellular stimulation 100-150 mm from soma along the apical dendrite (Fig. 1a,c). After obtaining a stable baseline measure of the EPSP waveform, EPSPs were repeatedly paired to postsynaptic action potentials (APs) evoked by brief somatic current injection with a delay of 3-8 ms ( Fig. 1c; see Methods section for more details on the spike-timing-dependent long-term potentiation (tLTP) induction protocol). After pairing, EPSPs were recorded for up to 30 min to measure changes in EPSP slope. This protocol elicits robust tLTP in L5 pyramidal neurons of mouse PrL-mPFC 23 . In L6 pyramidal neurons however, only a mild potentiation of EPSPs was induced (Fig. 1d,f,g; DEPSP slope: þ 7 ± 13%, n ¼ 8).
To activate nAChRs, nicotine (300 nM) was washed into the bath 2 min before and during the first 3 min of plasticity induction, which resulted in a clear postsynaptic depolarization (4.1 ± 1.0 mV, n ¼ 9, Fig. 1f bottom panel). Following these conditions, a strong lasting increase in EPSP slope was observed in response to pairing (Dslope: þ 65 ± 17%, n ¼ 9), which was significantly larger than that observed in control conditions (Fig. 1e-g; P ¼ 0.017). Omission of the postsynaptic AP during pairing in presence of nicotine did not result in a significant change in EPSP slope (Dslope: þ 11 ± 16%, n ¼ 4; paired samples t-test, P ¼ 0.791). These results show that in contrast to L5, acute exposure to nicotine at smoking-relevant concentrations facilitates tLTP of layer 6 pyramidal neuron synapses.
Endogenous ACh modulates STDP with layer specificity. Cortical nAChRs are activated by endogenous cholinergic projections from the BF [25][26][27] . To specifically activate cholinergic projections we used transgenic mice that expressed channel rhodopsin (ChR2) in Chat-positive neurons 28 , which allows for light-evoked ACh release from cholinergic terminals. We have previously shown that in L2/3, nAChRs are expressed predominantly by interneurons and only by a small fraction of pyramidal neurons 11 . Consistent with these reports, we observed nAChR-mediated currents in response to light in all L2/3 non-FS interneurons (10 out of 10) and in only a few L2/3 pyramidal neurons (2 out of 15; Fig. 2a). Conversely, in L6, light-evoked responses were observed in all pyramidal neurons. These currents were blocked by DHbE ( Fig. 2b; aCSF: 30.0±5.5 pA versus DHbE: 8.8 ± 2.7 pA, n ¼ 5, paired t-test, P ¼ 0.009), consistent with earlier reports that cholinergic fibers activate postsynaptic b2-containing nAChRs in L6 pyramidal neurons 29,30 .
To test whether the layer-and cell type-specific modulation of neuronal excitability by endogenous ACh results in layer-specific modulation of mPFC spike timing-dependent plasticity (STDP) rules, we performed tLTP experiments in L2/3 and L6 with lightevoked ACh release during plasticity induction. BF cholinergic neurons fire in short bursts in vivo during wakefulness 31,32 . To mimic naturalistic firing patterns of these neurons, blue light pulses of 10 ms each were delivered at 25 Hz (Fig. 2c), which is within the range of natural firing frequencies of BF cholinergic neurons observed in vivo 31 . In L2/3, EPSP þ AP pairing without light resulted in a robust tLTP ( þ 46±20%, n ¼ 12). However, endogenous ACh, released with a burst of blue light pulses preceding EPSP þ AP pairing, prevented tLTP ( À 7±15%, n ¼ 12; Independent samples t-test: P ¼ 0.044; Fig. 2d-f).
In L6, activation of cholinergic fibers evoked robust nAChR-mediated EPSPs (nAChR-EPSPs) with limited rundown or desensitization over the course of 50 trials with a 7 s trial interval (Fig. 2c). To investigate whether endogenous ACh could modulate tLTP, light-induced ACh release was evoked shortly before each EPSP þ AP pairing trial during plasticity induction, which caused the EPSP þ AP to approximately coincide with the peak of the nAChR-EPSP (Fig. 2g). EPSP þ AP pairing with coincident light-evoked ACh release resulted in tLTP ( þ 23 ± 10%, n ¼ 14), which was not seen in the control group of L6 pyramidal neurons from Chat-ChR2 mice that was not light-stimulated during pairing ( Fig. 2g-i; À 7 ± 11%, n ¼ 16; Mann-Whitney U-test, P ¼ 0.028). Altogether, these results show that brief cholinergic signals on naturalistic timescales 31 are sufficient to prevent tLTP of glutamatergic synapses in L2/3 pyramidal neurons, and to facilitate tLTP in L6 pyramidal neurons. Thus, endogenous cholinergic inputs modulate mPFC plasticity rules with layer specificity, having opposite effects in superficial versus deep cortical layers.
Facilitation of tLTP in L6 requires a5 nAChR subunits. Previous work has shown that L6 pyramidal neurons express nAChRs that contain b2 and a5 subunits 11,13,15 . However, a small fraction (20%) of L6 pyramidal neurons additionally express a7 nAChRs 11 . To test which type of nAChR mediates the effect of nicotine on tLTP, we made use of three strains of nAChR-knockout mice, each lacking a specific nAChR subunit. Nicotine (10 mM) was applied locally at somato-dendritic regions of the recorded cell from onset to offset of the pairing period (local application protocol II, Methods section), which resulted in a post-synaptic depolarization of similar magnitude during pairing as was observed on wash-in of nicotine (Figs 1f and 3a,b; wash: 4.1 ± 1.0 mV, n ¼ 9, versus local application: 5.1±1.2 mV, n ¼ 7). In L6 pyramidal neurons from wild-type (WT) animals, EPSP þ AP pairing in the presence of locally applied nicotine resulted in a significantly larger change in EPSP slope ( þ 40 ± 11%, n ¼ 7) than in control experiments where nicotine was not applied (Fig. 3b,f; À 2 ± 8%, n ¼ 8; one-way analysis of variance (ANOVA): F(1,13) ¼ 10.129, P ¼ 0.007).
a5-nAChRs are expressed at soma and dendrites of L6 neurons.
Since in the experiments above glutamatergic synaptic inputs were stimulated along the apical dendrite, and since the facilitation of tLTP depended on a5-containing nAChRs, we wondered whether a5-containing nAChRs are actually expressed at apical dendrites. To test this, local applications of ACh (1 mM, 10 s) were delivered either to the soma or to the apical dendrite at 200-300 mm distance from the soma of L6 pyramidal neurons ( Fig. 4a; local application protocol III, Methods section).
When a5-containing nAChRs are activated, they depolarize the cell membrane potential (Figs 1 and 3). Since a5-containing nAChRs are expressed along the apical dendrite and they enhance tLTP of glutamatergic synapses, activation of a5-containing nAChRs may facilitate AP propagation along dendrites and increase dendritic calcium influx. To test whether nicotine affects dendritic AP propagation, we investigated dendritic calcium signalling using two-photon calcium imaging. L6 pyramidal neurons were loaded with Alexa 594 (80 mM) to visualize neuronal morphology and the calcium indicator Fluo-4 (100 mM) to measure changes in dendritic calcium levels. Sections of primary apical dendrites were line scanned at a distance of 100-150 mm away from the soma towards pia (Fig. 5c) before, during and after nicotine application. To control for possible bleaching of the calcium indicator as a result of repeated line scanning, control experiments were performed, in which aCSF was washed-in instead of nicotine. Baseline dendritic calcium levels did not change significantly more in response to nicotine wash-in than aCSF (Median change nicotine: þ 0.53%, interquartile range ¼ 1.33%; Median change aCSF: 0.21%, interquartile range ¼ 2.49%, independent samples Mann-Whitney U-test: U (32) ¼ 118, P ¼ 0.719). However, in presence of nicotine, fluorescence transients following AP back-propagation were increased compared with aCSF, having both a greater amplitude (Fig. 5d,f; nicotine: 0.83 ± 0.11%DG/R, n ¼ 13; aCSF: 0.41±0.10%DG/R, n ¼ 16; P ¼ 0.008) and larger area (Fig. 5d,g, P ¼ 0.008). Fluorescence transients following bursts of somatic APs were increased in amplitude (Fig. 5e,h; nicotine: These results show that activation of nAChRs in L6 pyramidal neuron dendrites amplify dendritic calcium signals associated with dendritic AP propagation. Since dendritic calcium signalling is required for tLTP induction, enhanced dendritic calcium signals are likely the mechanism underlying the nicotine-induced facilitation of tLTP. nAChR distribution in human frontal and temporal cortex. Do any of the mechanisms of nicotinic modulation of tLTP occur in the neocortex of the human brain? The laminar pattern of nAChR modulation of mouse cortical pyramidal neurons has now been reported in many cortical areas, including prefrontal, motor, entorhinal and visual cortex 11,15,29,39 , showing that nAChRs more strongly excite pyramidal neurons of the deeper layers (L6 mostly) than those of the superficial layers. Autoradiography studies have shown a laminar distribution of nAChRs in human cortex as well 40 , but until now only cortical interneurons of the human frontal and temporal cortex were shown to express functional a7-containing and b2-containing nAChRs 41,42 . To test whether human cortical pyramidal neurons share a similar nAChR expression profile to rodents, we recorded from L2/3 and L6 pyramidal neurons (Fig. 6a) of human frontal and temporal cortex tissue resected during epilepsy surgery 43 (see Methods section and Table 1) and tested them for nAChR expression using direct applications of ACh (1 mM, 420 s; local application protocol II, in the presence of atropine) aimed at somato-dendritic regions of the cell (Fig. 6b). In L2/3, none of the recorded pyramidal neurons responded to ACh (0 out of 6 cells; Fig. 6a,b), similar to mouse L2/3 pyramidal neurons 11 . In L6, however, a subset of pyramidal neurons responded to ACh, with responses varying from modest 2-3 mV depolarizations to suprathreshold AP firing (Fig. 6a,c). The corresponding inward currents in response to ACh application were sensitive to the b2-containing nAChR antagonist DHbE (Fig. 6d), similar to mouse cortex 11 . These results suggest a similar laminar expression profile of nAChRs by pyramidal neurons as observed in the mouse brain.
Autoradiography studies have shown that nAChR expression depends on a person's smoking history [44][45][46] ; smoking increases nAChR levels in the brain, and after quitting smoking these return to pre-smoking levels 44,46 . It is however not known whether the up-regulation of nAChRs in smokers in fact leads to increased surface expression of nAChRs. To investigate this, we compared the data obtained from patients who smoked with that of non-smokers, pooling ex-smokers in our patient sample (three patients, all Z2 years of abstinence) with non-smokers. We found that in smokers, the distribution of nAChR-mediated postsynaptic potentials was significantly shifted towards higher amplitudes (non-smokers: n ¼ 67 (seventeen patients); smokers: n ¼ 14 (three patients); Kolmogorov-Smirnov test: P ¼ 0.004, Fig. 6e). This suggests that the increased abundance of nAChRs found in the brains of smokers, particularly in cortical L6, indeed leads to increased surface expression of functional nAChRs by pyramidal neurons in this layer.
Layer-specific modulation of tLTP in human cortex by nAChRs.
To test whether the laminar expression of nAChRs in human neocortex translates into layer-specific nicotinic modulation of synaptic plasticity, similar to mouse PFC, we performed tLTP experiments in L2/3 and L6 of human temporal and frontal cortex. In L2/3 pyramidal neurons, wash-in of ACh (1 mM, in presence of atropine (400 nM)) during pairing resulted in a complete blockade of tLTP compared with control conditions (Fig. 7a,b,d; Dslope ACh: À 5 ± 11%, n ¼ 9; Dslope aCSF: þ 44±16%, n ¼ 8; one-way ANOVA, P ¼ 0.019). These results indicate that tLTP is blocked by nAChR activation in cortical pyramidal neurons of superficial cortical layers, similar to L2/3 and L5 pyramidal neurons in mice 23,24 . In mouse L5, nicotine increased the threshold for tLTP by activation of presynaptic interneurons and correspondingly, L5 neurons displayed a mild hyperpolarization of the resting membrane potential following nicotine application 23 . In contrast to mouse L5 neurons, human L2/3 pyramidal neurons slowly depolarized with bath-application of ACh ( Fig. 7c; 2.5 ± 0.4 mV, n ¼ 9), suggesting that distinct mechanisms may be involved in nAChR modulation of tLTP in human L2/3 pyramidal neurons. Finally, we tested whether tLTP in human L6 pyramidal neurons is subject to modulation by nAChRs by performing plasticity experiments in the subpopulation of nAChR-expressing L6 pyramidal neurons. In control conditions, no tLTP was observed on average (Fig. 7e,f,h; Dslope: À 8±5%, n ¼ 7). Local application of ACh during pairing, which led to a modest but lasting depolarization ( Fig. 7g; 2.9±1.7 mV, n ¼ 6), resulted in an increase of EPSP slope (Dslope: þ 30.3 ± 16.6%, n ¼ 6) significantly larger than observed in control conditions (Fig. 7e,f,h; one-way ANOVA, P ¼ 0.036). Altogether, these results show that the laminar excitation of pyramidal neurons by nAChRs supports layer-specific modulation of human cortical STDP rules.

Discussion
In this study, we addressed the question whether endogenously released ACh controls plasticity of glutamatergic synapses in a layer-specific manner and what the underlying mechanisms are. We found that (1) in contrast to a suppression of plasticity in layer 5 (ref. 23), postsynaptic b2 and a5 subunit-containing nAChRs expressed by PFC L6 pyramidal neurons facilitate LTP of glutamatergic synapse strength. (2) Endogenous release of ACh can modulate cortical plasticity rules in a layer-dependent manner: tLTP is facilitated in L6 pyramidal neurons, but is suppressed in L2/3 pyramidal neurons. (3) a5 subunit-containing nAChRs are expressed at L6 pyramidal neuron dendrites, are activated by endogenous ACh and increase dendritic calcium influx and AP propagation, which is required for synaptic potentiation in these neurons. (4) In adult human neocortex, nAChRs are also expressed in a layer-dependent fashion in pyramidal neurons. (5) Similar mechanisms that result in layerspecific control of synaptic plasticity by nAChRs in mouse PFC also generate layer-specific modulation of synaptic potentiation in human neocortex. Together, these results show that the innervation of the prefrontal cortex by BF cholinergic neurons and the layer-dependent expression of nAChRs result in a layerspecific control of synaptic plasticity by endogenous ACh. This functional organization of the cortical cholinergic input system is most likely also in place in the adult human neocortex 47 .
Presynaptic nAChRs located on glutamatergic synaptic terminals have been well-known to directly modulate excitatory ARTICLE transmission and plasticity in several brain areas 16,17,19,[48][49][50] . In L5 of the PFC, presynaptic non-a7 nAChRs located on GABAergic interneurons alter synaptic plasticity of glutamatergic synapses on pyramidal neurons by reducing dendritic calcium influx during dendritic AP propagation 23,24 . In mouse hippocampus, timing-dependent plasticity can be modulated through a similar recruitment of inhibition by presynaptic nAChRs 51 . nAChR activity could bi-directionally modulate plasticity, and the sign of synaptic change was critically dependent on the timing and localization of nAChR activation. Stimulating a7-subunit-containing nAChRs with a local application of ACh to dendritic regions of the cell during plasticity induction boosts short-term into long-term plasticity 50,51 . If however, neighbouring interneurons were activated by nAChRs, the same protocol could no longer induce plasticity 51 . In deep layers of the entorhinal cortex, stimulation of non-a7 nAChRs also boosted short-term to LTP 39 , but neither mechanisms nor nAChR locations were identified. Here, we find that in PFC L6, dendritically located heteromeric nAChRs containing b2 and a5 subunits strongly augment synaptic potentiation of glutamatergic synapses by increasing dendritic calcium influx during dendritic AP propagation. Thus, in different layers of the PFC, dendritic calcium influx in pyramidal neurons is oppositely regulated by endogenous activation of nAChRs either located on the dendrites themselves, in the case of L6 pyramidal neurons, or on presynaptic GABAergic interneurons, in the case of the L5 circuitry. Given that in the mouse and human PFC L2/3 pyramidal neurons typically do not express nAChRs, in contrast to L2/3 interneurons 11,41 , activating cholinergic fibers in L2/3 most likely increases interneuron activity and inhibition of L2/3 pyramidal neuron dendrites, thereby inhibiting LTP, similar to L5. In a study of the dendritic properties of L6 pyramidal neurons of rat somatosensory cortex, it was found that the amplitude of back-propagating APs in apical dendrites of L6 neurons is particularly sensitive to the dendritic resting membrane potential 52 . In light of this, our finding that L6 pyramidal neurons express the tLTP-facilitating nAChRs along their dendrites is interesting, as dendritic nAChRs may well represent a source for such dendritic depolarizations, thereby acting as a physiological on/off switch for bAP enhancement and the induction of tLTP. Note that since the depolarization observed in a5 À / À animals in response to local application of ACh in dendritic regions is quite similar to that observed in WT animals (Fig. 4c), bAP-induced calcium signalling may be similarly amplified in a5 À / À animals. In that case, increased calcium influx through the a5-nAChRs expressed by WT animals may act to boost LTP directly.
In the PFC, L6 pyramidal neurons receive fast synaptic cholinergic transmission, but mediated by non-a7 nAChR containing b2 and possibly also a5 subunits 30 . We found here that optogenetic release of ACh at somatic, proximal dendritic and distal dendritic locations activates nAChRs with b2 and a5 subunits that enhance synaptic plasticity. Thus, glutamatergic synaptic plasticity may be augmented by fast cholinergic synapses on L6 pyramidal neurons. Given that distinct BF nuclei may preferentially target either superficial or deep layers of the PFC 7 , L1 and L2/3 interneurons may be innervated by a different population of BF cholinergic neurons than L6 pyramidal neurons. With the layer-specific and neuron-type specific distribution of nAChRs in the PFC 11 that can take part in fast synaptic cholinergic transmission 30,53 , a spatially detailed and millisecondscale temporal control of PFC glutamatergic and GABAergic signalling and plasticity by the BF cholinergic system is possible. Although the human neocortex shows structural similarities to the rodent neocortex, many striking differences in cellular and synaptic structure and function have been uncovered in recent years 43,[54][55][56][57] . Very little is known about whether cholinergic control of cortical processing in the human brain occurs through similar mechanisms as found in the rodent brain. In electron micrographs of the human temporal cortex, 67% of all varicosities on cholinergic axons formed identifiable synaptic specializations on spiny dendrites or spines 58 , which may suggest that fast cholinergic signalling could exist in human neocortex as well. Nicotinic AChRs are abundantly expressed in the human neocortex 40,45,46,59 and show a laminar distribution with the most dense staining in deep layers 40,46 . Cortical interneurons of human frontal and temporal cortex were shown to express functional a7-containing and b2-containing nAChRs 41,42 , but whether human pyramidal neurons express functional nAChRs was not known. We found here that similar to mouse PFC, nAChR expression in human pyramidal is layer-dependent. In response to local ACh application, pyramidal neurons in L2/3 did not show nAChR currents, whereas a substantial subset of L6 pyramidal neurons showed prominent inward currents carried by non-a7 nAChRs. Most likely, this layer-specific pattern of nAChR expression underlies the distinct effects of nAChR activation on glutamatergic synaptic plasticity in the human neocortex: suppression in superficial layer pyramidal neurons and augmentation in L6, similar to mouse PFC.
In tLTP experiments in human cortex where ACh was applied during plasticity induction, a reduction in EPSP slope was observed in the first minutes after EPSP þ AP pairing (Fig. 7a,e). It is known that in mice, nicotine enhances spontaneous and evoked inhibitory synaptic transmission to L5 pyramidal neurons. The apparent reduction in EPSP amplitude observed during recovery from exposure to ACh may therefore follow from such changes in inhibition/excitation ratio of PSPs evoked by extracellular stimulation. This reduction was most prominent in L2/3 pyramidal neurons; increased inhibition by ACh may therefore underlie the blockade of tLTP in these neurons, as was shown to be the case in mouse L5 pyramidal neurons 23 .
Layer 6 has a prominent role in cortical function. In visual area V1, layer 6 controls the gain of visually evoked activity in neurons of the upper layers 60 . This gain modulation depends on intracortical projections from L6 pyramidal neurons to superficial layers as well as projections to the thalamus [60][61][62] . PFC L6 pyramidal cells also connect to thalamic nuclei 63 and play a role in attention and top-down control 64 . How glutamatergic synaptic plasticity in L5 and L6 relates to attention performance and top-down control in mice is not understood at this point. Nevertheless, fast ACh signalling in the PFC at sub-second time scales relevant for nAChR activation during attending and detecting sensory cues 65 is directly involved in cognitive processing and mediates a shift from monitoring for cues towards the generation of a cue-directed response 66 . It is likely these cognitive processes depend on a balanced laminar control of PFC function by ACh.

Methods
Human neocortical brain tissue. All procedures on human tissue were performed with the approval of the Medical Ethical Committee of the VU University Medical Centre and in accordance with Dutch license procedures and the declaration of Helsinki. Human anterior medial temporal cortex and frontal cortex tissue that had to be removed for the surgical treatment of deeper brain structures was obtained with written informed consent of the patients before surgery. Non-pathological neocortical tissue showing no abnormalities on preoperative MRI was obtained from a total of 33 patients (32 adults (17 females, 16 males, aged 19-55 years) and one 9 year-old male), operated for medial temporal lobe epilepsy (15 cases), to remove hippocampal tumours (2 cases), cavernomas (4 cases) or for other reasons (12 cases). Our sample of human patients contained 7 smokers (11.5 ± 3.7 pack years), 23 non-smokers and 3 ex-smokers in abstinence for Z2 years (Table 1).
Human and mouse neocortical brain slice preparation. Human brain slices were prepared following the same routines as described previously 43 26 NaHCO 3 and 10 glucose (solution referred to as artificial cerebrospinal fluid (aCSF) throughout this paper). Here they were stored for B30 min at 34°C and subsequently for at least 1 h at room temperature before recording. All solutions were continuously bubbled with carbogen gas (95% O 2 , 5% CO 2 ), and had an osmolarity of B300 mOsm. All animal experimental procedures were approved by the VU University's Animal Experimentation Ethics Committee and were in accordance with institutional and Dutch license procedures. Mouse brain slices were prepared from P19-35 male or female C57BL/6 mice (referred to as WT throughout this paper), from mice lacking either a7-nAChR subunits (a7 À / À ), b2-nAChR subunits (b2 À / À ), or a5-nAChR subunits (a5 À / À ), or from Chat-ChR(N6) or Chat-Cre/Ai32 mice. Following decapitation, the brain was swiftly removed from the skull and placed in ice-cold slicing solution containing (in mM): 125 NaCl, 3 KCl, 1.25 NaH 2 PO 4 , 3 MgSO 4 , 1 CaCl 2 , 26 NaHCO 3 and 10 glucose. Coronal slices (350 mm) of mPFC were then cut and transferred into holding chambers and allowed to recover in aCSF for at least an hour.
Electrophysiology in human and mouse neocortical slices. Following recovery, slices were placed in a recording chamber and perfused with aCSF (3-4 ml min À 1 , 31-34°C). Layer 6 and layer 2/3 pyramidal neurons in human and mouse tissue were identified with oblique illumination or differential interference contrast microscopy. All experiments in mouse tissue were performed in the prelimbic area of mPFC. Whole-cell patch-clamp recordings were then made using standard borosilicate glass pipettes with fire-polished tips (4.0-6.0 MO resistance) filled with intracellular solution containing (mM): 110 K-gluconate; 10 KCl; 10 HEPES; 10 K 2 Phosphocreatine; 4 ATP-Mg; 0.4 GTP, biocytin 5 mg ml À 1 (pH adjusted with KOH to 7.3; 280-290 mOsm). Recordings were made using MultiClamp 700 A/B amplifiers (Axon Instruments, CA, USA), sampling at 10 kHz and low-pass filtering at 3-4 kHz. Recordings were digitized with an Axon Digidata 1440A and acquired using pClamp software (Axon). After experiments were completed, slices were stored in 4% PFA for subsequent neuronal visualization and reconstruction as described in detail in Mohan et al 57 . Spike timing-dependent plasticity. Spike timing-dependent plasticity experiments were performed using procedures described previously 23,43 . Excitatory postsynaptic potentials (EPSPs) were evoked every 7 s (0.14 Hz) using bipolar stimulating electrodes in glass pipettes filled with aCSF positioned B100-150 mm along the cell's apical dendrite (Fig. 1a). Duration (50 ms) and amplitude (40- . The slope of the initial 2 ms of the EPSP was taken as measure of EPSP strength. Change in synaptic strength was defined as percent change in EPSP slope 25-35 min. after onset of pairing relative to baseline. In case recordings lasted less than 20 min. after pairing, the whole post-pairing period (415 min after pairing to end) was compared with baseline (9 out of 178 included plasticity experiments). Cell input resistance was monitored by applying a hyperpolarising pulse at the end of each sweep ( À 30 pA in mouse and human L6 neurons, À 100 pA in human L2/3 neurons, 500 ms duration). After pairing, membrane potential was returned to approximate baseline value by modest current injection. Criteria for inclusion of recordings in STDP dataset were: (1) baseline resting membrane potential o À 60 mV, (2) smooth rise of EPSP and clear separation from stimulation artefact, (3) stable baseline EPSP slope, (4) less than 30% change in input resistance, (5) no AP-firing evoked by extracellular stimulation in post-pairing period. Two cases of extreme EPSP rundown (slope o20% of baseline) were excluded from analysis.
Light-evoked endogenous ACh release. Endogenous ACh release in Chat-Cre/ Ai32 and Chat-ChR(N6) mice was evoked from prefrontal cholinergic fibers with pulses of blue light (470 nm) using a DC4100 4-channel LED-driver (Thorlabs, Newton, NJ). tLTP experiments with light-evoked endogenous ACh release ( Fig. 2b-g) were performed in L2/3 and L6 pyramidal neurons of Chat-ChR2(N6) mice. Only L6 neurons that depolarized in response to a test pulse of blue light (10 ms duration) were included. The same tLTP protocol as described above was used in these experiments, with the exception that short bursts of light pulses (10 ms duration, 25 Hz) were given before each EPSP þ AP pairing (L2/3: five light pulses, starting 200 ms before; L6: 2 pulses, starting 80 ms before). In L6 pyramidal neurons, this caused the EPSP þ AP pair to approximately coincide with the peak of the light-evoked nAChR-mediated depolarization. To test whether cholinergic fibers target dendritic nAChRs (Fig. 4d), ChR2 was activated along the dendrites of layer 6 pyramidal neurons of Chat-Cre/Ai32 mice using an insulated optic fiber (core/cladding Ø 50/125 mm) held within a glass pipette (tip Ø 150 mm). To determine how spatially restricted ChR2 excitation was with optic fiber delivery of light, light-evoked ChR2 currents in ChR2-positive cells were recorded with the light spot at different distances from the cell. These measurements showed that there was a steep drop off of light-induced ChR2 currents when moving the light spot away from the cell. At 400 mm distance from the cell, less than 10% of the ChR2 current remained (data not shown).
Local application of nAChR agonists. Locally applied nAChR agonists were dissolved in aCSF including atropine (400 nM), loaded into glass pipettes and locally applied to neurons by pressure ejection. Three distinct methods of local application were employed in this study. Local application protocol I: ACh (1 mM) was applied for 10 ms using a custom built pulse generator attached to a pressure valve. Local application pipettes had a tip opening of B1 mm and were positioned B30 mm lateral from soma. Local application protocol II: ACh (1 mM) or nicotine (10 mM) was applied by syringe connected to a local application pipette, continuously at 30-40 mbar pressure for 10-350 s (durations vary per experiment and are specified in main text). Local application pipettes had a tip opening of B2 mm and were positioned B80 mm lateral from soma. Local application protocol III: ACh (1 mM) was applied for 10 s using a Picospritzer III (General Valve Corporation, Fairfield, NJ). Local application pipettes had a tip opening of B1 mm and were positioned B30 mm lateral from soma, or 200-300 mm away in the direction of pia, along the apical dendrite. Local application of the fluorescent dye Alexa Fluor 488 showed that the radius of the spread of this type of application in the slice tissue was around 40 to 50 mm. In tLTP experiments on human L6 pyramidal neurons, neurons were categorized as nAChR-positive if the response to ACh using local application protocol I/II was larger than twice the baseline s.d., orin tLTP þ ACh experiments-if neurons depolarized more than 2 mV in the initial phase of plasticity induction in response to local ACh application.
Two-photon Ca 2 þ imaging. Fluorescent dyes Alexa Fluor 594 (80 mM; Invitrogen) and Fluo-4 (100 mM; Invitrogen) were added to the intracellular solution to visualize morphology and measure [Ca 2 þ ] i changes respectively within the dendrites. After establishing whole-cell configuration in layer 6 pyramidal neurons, dyes were allowed to diffuse into the dendritic tree for 20 min before imaging. Single APs or bursts of 3 APs (40 Hz) were triggered by somatic current injection (1-2 nA) to induce back-propagating APs (bAPs). bAP-induced Ca 2 þ influx was assessed by the fluorescence change in the Fluo-4 signal relative to the corresponding constant Alexa Fluor 594 signal after background was subtracted from each signal 69 . Fluorescence was measured using a LEICA RS2 two-photon laser scanning microscope with a Â 40 (0.8 numerical aperture (NA)) or Â 63 (0.9 NA) water-immersion objective and a Ti:Sapphire laser tuned to 830 nm excitation at a bidirectional scanning frequency of 8 kHz. Line scans (500 ms duration, 8 bit signal) synchronized with AP stimulation were made at a dendritic region of interest (ROI) 100-200 mm from soma. To assess the effect of nicotine on dendritic bAPs in layer 6 neurons, nicotine (10 mM) was bath applied and identical stimulus protocols and line scans were repeated. Line scans were repeated 3-6 times per stimulus protocol per ROI and were averaged for analysis. Amplitude (mean %DG/R within 50 ms of (last) AP) and area (integral of trace (%DG/R*ms) from (first) AP to end of line scan (total window: 420 ms)) of fluorescence signal running average were calculated offline. Fluorescence signals with 40.5%DG/R baseline s.d. were excluded from the analysis.
Analysis and statistics. All raw data was analysed using custom Matlab scripts (R2009a, MathWorks) or Clampfit 10.2. Statistical analysis was performed using IBM SPSS statistics 21. Data were tested for normality using the Shapiro-Wilk test.
In case of a significant deviation from normal distribution, non-parametric statistical tests were used. Otherwise, the appropriate parametric statistical test as mentioned in main text was used. In all statistical comparisons, Po0.05 was taken as level of significance.
Data availability. The data that support the findings of this study are available from the corresponding author on request.